In 2000, Mandyam Srinivasan, Shaowu Zhang, Monika Altwein and Jürgen Tautz published a paper in Science showing that bees estimated distance travelled during flights by “the extent to which the image of the world has moved on the eye”. In January 2016, I revisited this paper with Mandyam Srinivasan, to discuss its impact and relevance in the context of current research on the topic.
Date of interview: 17th January 2016
Hari Sridhar: Have you ever read the paper completely after it was published?
Mandyam Srinivasan: I suppose I haven’t read the paper completely since the time it was published, but there were several sorts of things that it lead to – experiments one could do based on what we found in this paper. And so we have continued work along those lines.
One of the things that this study suggested was that you could scale down large-range flights, which happen outdoors in natural environments, to the lab. You can simulate a long flight using a short tunnel. With this, you can do interesting things, like study how bees do what is known as path integration – when a bee goes along a tortuous route, where it makes turns left and right. That’s hard to setup in an outdoor environment. You cannot force bees to take a particular route outdoors, but indoors you can set that up – make bees go along a particular trajectory and then see how they integrate the distances they have travelled, till they come up with a final vector that shows the distance and the direction to the food source. So these kinds of things we started to do after this paper.
HS: Apart from the research it lead to in your own lab, can you give us a sense of research on this topic before and after this paper, to get some idea of what kind of impact this paper has had.
MS: I guess people have, since then, done similar kinds of experiments with other species of insects. They have worked with stingless bees, and I think also with wasps – though I am not totally sure of that. But certainly with stingless bees, where they have demonstrated similar behaviour. And probably a little more significant is the fact that a number of robotics groups have started to use this to enable aircraft systems to measure distance flown. In the absence of GPS and things like that, can we use vision to do our odometry? That question has been there for a while – I wouldn’t say we are the first ones to suggest that approach – but certainly the fact that animals do it has encouraged a lot more people to use that approach, ourselves included. So we are now putting these odometric systems into our flying robots as well. And it’s turning out to be quite useful.
HS: At the time when you published the paper, did you have some expectation of the kind of impact it might have? How does what has happened compare to that expectation?
MS: I really don’t know, Hari. It’s hard for me to tell. I have to go back and look at the citations. But people are starting to use the tunnel. We were probably the first ones to fly bees in tunnels, and since then a large number of people have started to fly bees and other insects in tunnels. That’s probably one of the main impacts – the general idea of scaling down these experiments in the lab, making them fly in the tunnel and looking at all these behaviours that they do – controlling their flight speed, negotiating narrow gaps, measuring distance travelled and so on. That’s probably largely due to the work that came out from our lab, I would say. But there are some interesting puzzles in it that still bother me and which still haven’t been solved, so maybe we could talk a little about that.
MVS: So one of the things we said was – okay, a bee flies a certain distance to a food source, it comes back and reports to its nest mates, through the dance, how far it thinks it has travelled, based on the optic flow it experienced. The optic flow varies depending on whether the environment was tight and close, or more open, with objects further away, right? But that does not matter because another bee that observes the dance and flies to find the food source will take the same route – it will experience the same environment – with the same optic flow. In other words, it doesn’t matter if your yardstick isn’t correct as long as everyone is using the same yardstick, right? That’s one thing. But there is another potential problem – if the bees are using the ground to measure optic flow, then the height they are flying above the ground matters – when you are flying very low, optic flow is high, and when you are flying high, it is low. Either everybody is flying at the same height above the ground, or they have to have some mechanism for knowing how high they are. This is still a mystery – we don’t know how they do it. If you fly in a very dense forest, then there is a lot of optic flow coming from the sides, because you are passing lots of tree trunks and things like that. The ground may not be very important then. But if you are flying in an open meadow, then all you have is the ground right? That is your only source of optic flow. Then you have to somehow regulate the height, or know it and account for it. How they do that is still not known. We tried some experiments to see whether they are looking mainly at the floor or the sides but they weren’t very conclusive.
HS: When you read this paper now, what strikes you about it?
MVS: It is interesting. Usually, I get very worried when I read my own papers. I feel – why did I study this, or it doesn’t make sense, or I could have expressed it better. But this one, strangely, doesn’t seem like that. At least, I cannot find anything I am particularly embarrassed about! One thing of course is that once the work has been done and published, everything seems quite obvious, right? Some of these calculations I have done here, they almost seem a little superfluous. Someone could look at this and say they could surely work that out for themselves; that it is a bit redundant. But there are also some things I feel, maybe, we could have elaborated upon, if there was more space. For example, explaining we are not the first one to investigate the nature of the honeybee odometer. Karl von Frisch, the Nobel laureate, had already looked into that, and had done some very interesting experiments. He did outdoor experiments, where he had bees flying from a hive to a feeder he had placed a long distance away – 500 metres away – and then measured the dance. It turned out that the bees were indicating the correct distance. Then he put small weights on these bees, made them fly the same distance again and found they signalled a much longer distance. From this he concluded that the bees must be estimating distance based on energy consumed. But we think what was happening there was that the bees with weights were flying closer to the ground and therefore experiencing much greater optic flow. This has not really been tested yet, so that’s something that someone should probably try and do. It would be interesting to see what they find. Again, these things are hard to do in the field, you know, because it is hard to measure the flight of bees outdoors.
HS: Is this something you wanted to talk about in the paper but didn’t?
MVS: Yes, I would have liked to talk about this, but space was a limiting factor.
HS: What else do you recall about your experiments, when you read this paper now?
MVS: I was so sure these experiments would not work, because we had never worked with bee dances before. We had done a lot of other experiments with bees – training them in a Y-maze to do pattern recognition and various kinds of discrimination tasks. All of that we had done in the past. But we had never actually looked at the dance. So it’s really, I suppose, beginner’s luck. One of our collaborators in Germany, and the senior author on the paper – Jürgen Tautz – sent to us a student – Monika Altwein – who, I think, was a senior undergraduate at that time, to do a little project in our lab. Part of her project, actually, was to show us how to film and analyse the bee dance. That was how we ended up studying dance behaviour for the first time. So it was really nice that we could get a paper published in a journal as big as Science! Very fortunate. In fact, I was not at all sure if we could fool the bees in such a comprehensive manner, just by putting them inside a slightly unusual environment. That’s why I was convinced that the experiment would not work. The first few times we did the experiments, we couldn’t believe that it actually worked. But I think that ours is certainly not the full explanation; I don’t think it is the full story. I find it hard to believe that if a bee flies, for example, distances greater than 10 kilometres in search of food, it relies purely on optic flow. It must feel tired, it must feel exhausted, its crop would be almost empty, and it would feel a sense of fatigue – all these would also indicate distance in some way, I think. Maybe for short to medium flights, optic flow is the only cue, but for longer flights I think other cues will be required.
HS: Do you continue to work with your co-authors on this paper?
MVS: No, not anymore actually. We sort of drifted apart. Zhang Shaowu was a colleague of mine – in fact, I sort of appointed him to our department – and we have done a lot of lovely work together. I owe him a lot really, for all the stuff that came out during that period. We were working together in Canberra for a long time, at ANU, but when I moved to Brisbane, he stayed behind in Canberra and continued his work. He has retired now, but he still continues to work, to collaborate with a lot of universities in China. And he continued the collaboration with Jürgen Tautz too, for a while, but I’m not sure what has happened recently. I’m not sure what Monika is doing now. I’m not even sure if she did a PhD. I keep in touch with Jürgen Tautz once in a while, exchanging New Year greetings and things like that, but I don’t work with him anymore.
HS: What about the experimental setup and the colonies – do you know what’s happened to them?
MVS: The colonies we used would have died by now and there will be new generations. But that facility, I think, was still being used, until recently, by a colleague of mine, a guy who is a molecular neurobiologist. His name is Ryszard Maleszka, and he was interested in learning and memory in bees, and looking at the molecular basis of learning. He was using those colonies for a while. I don’t know what’s happened now.
HS: This facility was in the university campus?
MVS: Yes, there were a couple of different areas. One was an all-weather bee flight facility – a greenhouse which was converted to a flight facility. The idea was that we could do experiments with bee flight indoors, independent of the weather outside. We had hives which were located inside the greenhouse against the wall, with a small opening to let the bees go outside and get their food and everything, and also an opening on the inside, which we opened from time to time to let the bees into our arena for training and experiments. When I moved to Brisbane, they built another facility like that for us, in the rooftop of one of the buildings. We also had another sort of a small field station, where we could do outdoor experiments with bees. Both these facilities were used by others, after I left, till recently. Sadly, I think that insect vision programme over there is drawing to a close, because all the people there are getting old and close to retirement. Just like me!
HS: Do you think your writing style has changed from the time you wrote this paper?
MVS: Yes, I think it has gotten worse now! When I read this paper now, I am quite happy with the way the writing sounds.
HS: Did you do most of the writing for this paper?
MVS: Yes. As it turns out, in most cases, even if I’m not the senior author, I end up doing quite a lot of the writing because, quite often, students are relatively inexperienced. I try to encourage them a lot, of course, to write their own things, but sometimes there comes a point when, if you are trying to submit it to a journal, you have to make sure the language is reasonable. So I did have a fairly strong hand in the writing of this paper.
HS: It is very nicely written and easy to understand even for an outsider like me.
MVS: Yes, at least in the old days, it was important, if you were trying to publish in Nature or Science, to be as lucid as possible. Things have changed now. If you read the average Nature or Science paper nowadays, it’s so highly specialised, so technical, it’s very hard to follow. And there is a lot more material too; you are expected to put a lot more into it. A small idea, a small result, is no longer something that Nature or Science will even think about, unless it’s something really profound. And most often, nowadays, you need to have a strong molecular biology component, otherwise journals don’t even look at it. I think, if I submitted something like this now, it probably won’t get accepted.
HS: Did this paper have an electronic supplement?
MVS: No. As far as I can recall, everything went into the main paper. I think the whole concept of supplementary material came a little bit later. And now of course, 90% is in the supplementary material. The actual paper is like the tip of the iceberg.
HS: Do you remember if your paper changed a lot from the first draft?
MVS: I don’t think so. Actually, we were surprised it went through very smoothly. I think we got it back with fairly minor corrections, you know. In fact, I got the good news when I was in Bangalore on a visit; so that was nice.
HS: Given the theoretical and technological advancements in your field, if you were to redo these experiments today, would you change anything?
MVS: Okay, when filming the dance behaviour of bees on the hives, you have to be careful not to shine any directional light on them, because that can distract their dancing. In fact, they assume that light to be the sun, get disoriented and confused, and start to use that as a bearing. Of course, without light we can’t do any filming, so we tried our best to provide a diffuse source of illumination. We did some controls and there didn’t seem to be any steady bias, so we went ahead and used it. But if I were to do the experiments again today, I would probably use infrared illumination, which the bees don’t see. The cleanest way to do this experiment is to have a totally dark room, shine infra-red light on the dancing bees – which the bees cannot see – and then use an infrared sensitive camera to film it.
The other thing I would change is the analysis of the dance behaviour. For this paper, we digitised and analysed the dance manually, step-by-step. It was very tedious and painstaking work, trying to track the position of these bees over a huge number of video frames. But since then, an engineering student from the University of Maryland US, who came to do a little stint with us for a few months, has developed a nice algorithm for automatically tracking bees while they were dancing. With these automated bee trackers it is a lot easier to analyse the dances and get a lot more data. So certainly that would have been a very useful thing.
HS: And in terms of experimental design – would you change anything if you were to repeat the experiments today?
MVS: No, I can’t think of any way to improve upon it.
HS: What would you say are the main takeaways from the study?
MVS: To my mind, it proves, fairly conclusively, that it is optic flow that bees are using to measure distance travelled. That really is the main thing. And it also calibrates that, in terms of units of optic flow per waggle dance duration. It’s a new way of calibrating the odometer, because previously it was all in terms of distance travelled. But now, we are saying the fundamental unit the bee is measuring is not the distance, but the amount of optic flow. For me, looking back on it and seeing what Von Frisch did – he did those very nice, clever weighting experiments, but, as I described earlier, it seems like he might have been misled. It worries me that, by the same token, we could have missed something very important, which someone else later might come and point out.
HS: But it hasn’t happened so far?
MVS: No, It hasn’t happened.
HS: And you haven’t discovered anything problematic, yourself?
MVS: I haven’t thought of anything. I mean, like I said, there is still a lot which is unexplained here, but I haven’t seen anything that contradicts, or finds a completely different explanation for, our results. Except for very long distance flights. I don’t know if anyone has really found other cues that bees could be using when they fly large distances.
HS: And is the calibration for waggle duration in terms of optic flow units – “17.7 degrees of image motion per millisecond of waggle” – still undisputed?
MVS: I would say so, but I am not sure. I have to go back and check. That’s a good point. Certainly, our subsequent experiments with the polarised light give the same numbers, and I know people have done something similar with stingless bees. But I will have to go back and take a look to say for sure. That’s a good point.
HS: What kind of impact has this paper had on your career?
MVS: Well, I was being considered for election to the Royal Society around the time this paper got published. So maybe it made a difference to my election as a fellow. Also, I had applied for what’s called a Federation fellowship, one of those nice senior professorial fellowships, which are highly competitive in Australia. And that came through as well. So I do think it helped my career quite a bit. And to think that the work itself was almost completely accidental, a result of just sitting and musing and saying – what would happen, you know, if something like this happened etc. It certainly didn’t come from a grant we wrote up; it was just something on the side.
HS: Does this paper come up in conversations even now?
MVS: Quite often. At least people in the field seem to remember it and know it quite well. Although, I must say, I don’t know how much of an impact it has really had on the science.
HS: You don’t keep track of where it’s being cited?
MVS: No, I haven’t really looked at the papers that have cited it. I mean, I suppose I could go into Google and check that.
HS: As of today, it’s been cited 280 times
MVS: Okay, so it’s not a very high impact paper. It’s been around for a while. I haven’t looked at the papers in detail to see in what context they have cited. Maybe I should do that sometime. Maybe when I retire!
HS: Is that because the papers that cite this are somewhat outside your main research interests today?
MVS: Maybe. I have seen a few engineering papers cite it, just to say – hey look, animals do this too you know. They will say they are going to get their robots to use visual odometry to measure distances travelled or something and by the way, animals do it too. Just to add strength to the paper. That sort of thing.
HS: In concluding your paper you say “In the future it should be possible to study mechanisms of navigation and path integration conveniently, and under controlled conditions, by training bees to fly through short tunnels arranged in various configurations”. Has this happened?
MVS: Yes, we have added a third dimension, to see what happens when bees fly a stretch horizontally and then fly vertically upwards, to get to a food source. The interesting question there was whether they will measure the optic flow rate independently of the direction in which the image is moving in the eye? Because when they are flying horizontally, the image is moving from front to back in the eye, but when they go up the tunnel – bees fly up like helicopters – the image flow is vertically downwards. We wanted to know whether they would integrate horizontal and vertical distances in the same way, and it turns out that they do. What is even more interesting is that they don’t sum these two distances vectorially, but instead provide the total scalar distance travelled, i.e. the perimeter distance. Very interesting; I don’t know why they do that. Even in the polarised light experiments, where we simulated forward flight and then flight in a perpendicular direction, the bees gave the direction as the vector of the sum of those two vectors, but the length was the total. Why they do this is still an unsolved enigmatic puzzle. Some people say – okay, if a bee is forced to take a detour around a hill or something, to go to a particular food source, then the physical distance that the bee needs to travel is not the vector but the perimeter distance, right? So if the bee comes back and gives this full distance that needs to be travelled, that is useful information for the other bees, which are contemplating whether or not they should go to that particular food site. If the bees are comparing two different potential food sources to decide which one to go to, it might help to get an idea of the true distance and not the short cut distance. That’s the best explanation so far, but it is still a bit of a puzzle.
HS: Last question – in general, do you ever find yourself going back to the papers you have published?
MVS: If I am working on something that’s based on what we have done in the past I would certainly go back to check exactly what we have said and so on. The other time I do it is when someone emails me with a question about it. That happens quite often- usually, people asking about methods. So then I go back and read the paper. Also, when writing reviews, I often have to go back and read my earlier papers.